Response to classification numbers of vegetation types on correlative coefficients among landscape metrics

GONG Jian-Zhou,XIA Bei-Cheng

DOI: https://doi.org/10.3321/j.issn:1000-0933.2007.10.015

2007-01-01

Abstract:It is important to analyze characteristics of landscape pattern indices for landscape ecological research, but its development and application in the field has been limited because of complicated relationships among the indices. There are more than 100 indices used to describe characteristics of landscape system. In this research 28 indices were analyzed, which were defined in three types of landscape indices (area/perimeter/density, shape, and contagion). The basic characteristics of a regional landscape were based on maps of vegetation fraction grading, which were translated from TM remote sensing images of Guangzhou on Oct. 22, 2005. This research was performed in the city center (8 districts) of Guangzhou. All landscape indices were calculated with the software Fragstats3.3. At the level of class metrics, correlative relationships among indices, which were in the same type of classes, were explored firstly. Then, the response of the correlative relationships was studied under different classification systems of vegetation fraction for the same landscape, and seven kinds of classification systems of vegetation fraction were designed, namely 2, 3, 4, 5, 6, 9 and 17. The results of correlative relationships were as following: relationship between every index pair was obvious for almost all indices, but there were great difference among different index pairs. For the type of area/perimeter/density metrics the relationships were most significant, and for the type of shape metrics indices were strong independent from each other. Some indices, such as PD, have their different ecological meaning at the landscape level and at the class metrics level. This research got a conclusion, which was different from literatures, by calculating correlative relationship between PD and AREA_MN. It revealed that it should be important to choose suitable indices for the goal of research at the class level and the landscape level. The results on sensitivity of those indices to number of types were analyzed as following: most of indices were very sensitive to classification number of landscape types, and the response degree and tendency were different among different metrics. The response tendencies of 117 index pairs were divided into three groups according to their action under different numbers of landscape types, such as non-sensitiveness, monotonous increasing and complex, and the number of index pairs of each group was 12, 31, and 74, respectively. The response to the number of landscape types indicated that classification numbers 4 and 5 were most sensitive. Another result showed that response difference stood among different types of class metrics, and indices about Area/Perimeter/Density metrics was non-sensitive with changing of classification numbers of types. The metrics about shape of landscape was the most sensitive. Additionally, the results revealed that the correlative relationship of indices were related to spatial characteristics of position.